EE images from separate rolls but collected on the same date must be pre-processed separately

```
batch.EE_pre_process_images(
api_key,
project_name = 'GTNP',
bounds = (-110.82, 43.75, -110.77, 43.70),
ee_project_name = 'VBRW0',
year = 1967,
month = 8,
day = 17,
pixel_pitch = None,
buffer_m = 2000,
threshold_px = 50,
missing_proxy = None,
download_images = True,
template_parent_dir = '../../../hipp/examples/fiducial_proxy_detection/input_data/fiducials/ee/block',
output_directory = './',
ee_query_max_results = 100,
)
```

Returns 14 images collected on the same date, with the same camera.

```
AR1VBRW00010110
AR1VBRW00010111
AR1VBRW00010112
AR1VBRW00010113
AR1VBRW00010136
AR1VBRW00010137
AR1VBRW00010138
AR1VBRW00010139
AR1VBRW00010156
AR1VBRW00010157
AR1VBRW00010158
AR1VBRW00010159
AR1VBRW00010160
AR1VBRW00020010
```

The last image is part of a different roll (AR1VBRW000**2**0010) and should be pre-processed separately.

Need to dive into the logic and see where best to handle this, either in HIPP or HSfM. Returning only one images from a roll is an edge case that should be handled gracefully.

Start at the batch.EE_pre_process_images entry point and trace how returned Earth Engine images are grouped before preprocessing. Compare the roll identifiers in the example, including the singleton roll, and determine whether the change belongs in HIPP or HSfM; done means images from different rolls are processed separately without breaking single-image rolls.

Written by the indexing model from the issue text.